3752
J. Phys. Chem. B 2001, 105, 3752-3758
Characterization of Various Oxygen Species on an Oxide Surface: RuO2(110)† Y. D. Kim, A. P. Seitsonen,‡ S. Wendt, J. Wang, C. Fan, K. Jacobi, H. Over,* and G. Ertl Department of Physical Chemistry, Fritz-Haber-Institut der Max-Planck-Gesellschaft, Faradayweg 4-6, D-14195 Berlin, Germany ReceiVed: September 12, 2000; In Final Form: January 18, 2001
The stoichiometric RuO2(110) surface is terminated by bridge-coordinated oxygen atoms (Oβ) and by coordinatively unsaturated Ru (Rucus) atoms. Exposure to gaseous O2 leads to the formation of two additional surface species: a molecularly chemisorbed state (Oδ) bridging two neighboring Rucus atoms and weakly held O atoms (Oγ) in terminal position above the Rucus atoms. Characterization of the energetics and kinetics as well as structural, vibrational, and electronic properties is achieved by combined application of experimental (low-energy electron diffraction, high-resolution electron loss spectroscopy, thermal desorption spectroscopy) and theoretical (density functional theory) methods. The interplay between the different oxygen species accounts for the high sticking coefficient for dissociative adsorption as well as for the continuous restoration of the surface structure in the course of catalytic oxidation reactions.
1. Introduction The bond strength and the charge state of oxygen species on the catalyst’s surface are considered to govern the activity and selectivity of an oxidation catalyst. This makes a rich oxygen chemistry a necessary prerequisite for a versatile oxidation catalyst.1 Silver, for instance, is able to stabilize at least five different oxygen species at the surface, i.e., the molecular adsorption state, a chemisorbed atomic species, a dissolved O species, an embedded O species, and the oxidic O species. These O species reveal distinct chemical properties. For the methanol dehydrogenation reaction to formaldehyde over Ag, the strongly held embedded O species in the presence of dissolved O is important,2 while the chemisorbed species3 is suggested to determine the selectivity of the epoxidation reaction of ethylene. Several oxygen species have also been reported for ruthenium: the virtually inactive chemisorbed atomic O species,4 an oxidic species in the oxygen-rich phases,5 and a weakly held oxygen adsorbed on the oxygen-rich phase.6 The latter two species are quite active in the CO oxidation reaction.6,7 Lowenergy electron diffraction (LEED) and scanning tunneling microscopy (STM) investigations showed that the oxygen-rich phases on Ru(0001) consist of both (1×1)-O and RuO2(110) patches.5,8 The catalytically active parts of the oxygen-rich phase on Ru(0001) were identified with the (110) rutile face of RuO2,5,9 whose active centers are the coordinatively unsaturated Ru sites (Rucus atoms). The RuO2(110) surface is terminated by undercoordinated bridging oxygen atoms (Obr). Structural defects on the oxide surface do not play a significant role for the activity of RuO2, quite in contrast to the general view of active centers on oxide surfaces.10 Oxygen postexposure to the oxygen-rich phase at room temperature creates a weakly bound oxygen species with a dissociative sticking probability of 0.7. † Dedicated to Professor J. T. Yates on the occasion of his 65th birthday. Part of the special issue “John T. Yates, Jr. Festschrift”. * Corresponding author. Fax: +49-30-8413-5106. E-mail address:
[email protected]. URL: http://w3.rz-berlin.mpg.de/pc/pc.html. ‡ Also at INFM, Unita ` di Roma, Dipto. di Fisica, Universita` La Sapienza, P.’le A. Moro 2, I-00185 Roma, Italy, and IPP Garching, Boltzmannstr. 2, D-85748 Garching, Germany.
This O species desorbs at 450-500 K, and it efficiently catalyzes the CO oxidation even at room temperature.6 In this paper we review the various oxygen species on the RuO2(110) surface in comparison with those on Ru(0001) and present new data by applying the techniques of LEED, highresolution electron loss spectroscopy (HREELS), and density functional theory (DFT) calculations. Our combined study gives strong evidence that the weakly held oxygen species are adsorbed oxygen atoms which are located in terminal position above the coordinatively unsaturated Ru sites (Rucus) on RuO2(110). The vibrational characteristics of the RuO2(110) surface were studied by HREELS: two dominating losses at 69 and 103 meV are assigned to bridging O and on-top O, respectively. Furthermore our DFT calculations predict a chemisorbed molecular oxygen species which lies down on the RuO2(110) surface, bridging neighboring Rucus atoms. The presence of this O2 species is confirmed by thermal desorption spectroscopy and HREELS. The molecular oxygen species on RuO2(110) is considered to play a decisive role in the high dissociation probability of O2 over RuO2(110). 2. Experimental and Theoretical Details The LEED experiments were conducted in an ultrahigh vacuum chamber equipped with four-grid LEED optics, thermal desorption spectroscopy (TDS), and facilities for surface cleaning and characterization.11 The LEED intensity data were recorded at a sample temperature of 110 K, using a computerized, high-sensitivity CCD camera system. The analysis of the experimental LEED data used a multiple scattering program code that allows for the simultaneous and automated refinement of structural (and nonstructural) parameters.12,13 The agreement between calculated and experimental data was quantified by Pendry’s reliability factor RP.14 The HREELS measurements were performed in a second ultrahigh vacuum apparatus, which consists of two chambers. The upper chamber contains an argon ion gun, a quadrupole mass spectrometer, and a LEED optics. The lower chamber houses an HREELS spectrometer for recording the vibrational spectra. The HREEL spectra were all taken at a 60° angle of
10.1021/jp003213j CCC: $20.00 © 2001 American Chemical Society Published on Web 03/29/2001
Various Oxide Species on RuO2(110) incidence with respect to the surface normal and in specular geometry. The sample temperature was 83 K. The energy resolution was set to be 2.7 meV, which ensures a typical counting rate of 3 × 105 counts/s in the elastic peak. For the DFT calculations we employed the generalized gradient approximation of Perdew et al.15 for the exchangecorrelation functional and used ab initio pseudopotentials in the fully separable form.16 The surface was modeled by five double layers of RuO2(110) (supercell approach). Consecutive RuO2(110) slabs were separated by a vacuum region of about 16 Å. Calculations were performed using a (1 × 1) surface unit cell. Further details can be found in ref 7. To compute the vibrational frequencies, the position of the respective atom (consistent with the vibrational mode to be studied) was varied symmetrically around its equilibrium position, keeping all other atoms fixed at their equilibrium positions. The corresponding data points (in general five including the equilibrium position) were fitted to a parabolic function from which the vibrational frequency was derived. In both chambers (HREELS and LEED), the Ru(0001) sample was cleaned by argon ion bombardment at 1 keV followed by cycles of oxygen adsorption and thermal desorption in order to remove surface carbon. Final traces of oxygen were removed by flashing the surface to 1530 K, resulting in a sharp (1 × 1) LEED pattern. No impurity losses were observed in HREELS. The oxygen-rich ruthenium phase for the LEED experiments was produced by exposing a well-prepared single-crystal Ru(0001) to high doses of oxygen at an elevated sample temperature of 600 K. To reduce the oxygen background pressure in the ultrahigh vacuum chamber, oxygen was dosed through a glass capillary array doser (with channels 3 mm long and 10 µm wide, total transparency of 50%) about 1 mm away from the sample. In this way, the local oxygen pressure at the sample could be enhanced by a factor of about 100, thus allowing the oxygen partial pressure in the ultrahigh vacuum chamber to be kept below 10-5 mbar during dosing. Typical oxygen exposures (local pressure × time) for the preparation were 6 × 106 langmuirs. After the background pressure in the ultrahigh vacuum chamber has reached a value below 10-9 mbar, the sample was briefly annealed to 600 K in order to remove contamination by residual gas adsorption. The sample was then cooled to 100 K. The total amount of oxygen in the oxygen-rich Ru(0001) surface used in the LEED experiment was equivalent to about 6 monolayers, as estimated by a thermal desorption experiment. The corresponding LEED pattern indicates that the surface consists of patches of both Ru(0001)(1×1)-O and RuO2(110). The oxygen-rich Ru(0001) surface for the HREELS measurements was produced by exposing 1 × 107 langmuirs of O2 at 700 K. The glass capillary array doser was about 15 mm away from the sample, yielding an enhancement of about 30. This procedure resulted in a surface, which was completely covered by RuO2(110), as checked by LEED and HREELS. The weakly held oxygen (Oγ) was produced by exposing several langmuirs of O2 (for instance 5 langmuirs) to RuO2(110) at room temperature, while the molecular oxygen (Oδ) was prepared by exposing O2 to the sample at 100 K. Note that O overlayers on Ru(0001) are not able to stabilize a O2 species above 40 K. For the definition of the various O species on Ru(0001) and RuO2(110) we refer to the TD spectrum in Figure 1. The desorption state of the chemisorbed atomic oxygen species on Ru(0001) is indicated by (OR), while the oxidic oxygen of RuO2(110) is denoted as (Oβ).
J. Phys. Chem. B, Vol. 105, No. 18, 2001 3753
Figure 1. Thermal desorption spectra of oxygen on Ru(0001) and RuO2(110). The desorption states OR, Oβ, Oγ, and Oδ correspond to chemisorbed oxygen on Ru(0001), lattice oxygen of RuO2(110), weakly held oxygen on RuO2(110), and molecular oxygen on RuO2(110), respectively. The spectra of Oγ and Oδ are amplified by a factor of 5.
3. Results and Discussion 3.1. Surface Structure. Chemisorbed oxygen is able to form four ordered overlayer phases on Ru(0001), i.e., (2×2)-O, (2×1)-O, (2×2)-3O, and (1×1)-O. In all cases the oxygen atoms occupy hcp 3-fold hollow sites with the O-Ru bond length decreasing from 2.03 to 2.00 Å with increasing coverage.4 Increasing the O coverage beyond 2-3 monolayers leads to the epitaxial growth of RuO2(110) on Ru(0001) which may coexist with (1×1)-O on Ru(0001), depending on the sample temperature during preparation. By using NO2 instead of O2, O-rich phases with O contents of up to the equivalent of 5 monolayers on Ru(0001) can be produced without oxidizing the sample.9,17 These surface structures are characterized by a (1×1)-O overlayer and several monolayers of oxygen incorporated in the near-surface region.9 The as-grown RuO2(110) surface is terminated by a bridging oxygen atom Obr.8 In the bulk structure of RuO2 the Ru atoms are 6-fold coordinated to oxygen atoms, while the O atoms are coordinated to three Ru atoms in a planar sp2 hybridization. On the RuO2(110) surface two kinds of undercoordinated surface atoms are present, namely, the bridging oxygen atoms, which are coordinated only to two Ru atoms underneath (bond length: 1.94 Å), and the so-called Rucus atoms, i.e., 5-fold coordinated Ru atoms.8 The other Ru-O bond lengths are in the range of 1.90-2.03 Å. Noteworthy, DFT and LEED gave practically identical structural parameters for the RuO2(110) surface. To study the structural characteristics of the weakly held oxygen (Oγ), we monitored the intensity variation of LEED beams associated with the (1×1)-O and RuO2(110) phases during postdosing oxygen at room temperature. O2 exposure at room temperature does not change the appearance of the LEED pattern, and only the RuO2(110) related LEED intensities were affected. This implies that the weakly held oxygen atoms Oγ adsorb only on the oxide domains rather than on the (1×1)-O areas. Besides the absolute intensity change on O2 exposure, also the structure of the LEED IV curves altered. We therefore conclude on a specific (i.e., well-defined) adsorption site of Oγ. A detailed LEED analysis of Oγ-RuO2(110) related LEED IV curves disclosed the adsorption site of Oγ to be directly above the Rucus atom (cf. Figure 2) with an O-Ru bond distance of 1.85 Å. The agreement between experimental and calculated LEED IV curves for the optimized model structure of OγRuO2(110) is quantified by RP ) 0.35 (cf. Figure 3). The LEED analysis yields that 75 ( 20% of the Rucus atoms are occupied
3754 J. Phys. Chem. B, Vol. 105, No. 18, 2001
Kim et al. TABLE 1: Calculated DFT Binding Energies for the Systems O-Ru(0001) and RuO2(110)a
O species
binding energy (eV)
O species
binding energy (eV)
Ru(0001)-(2×2)-O hcp Ru(0001)-(2×1)-O bridge Ru(0001)-(2×2)-O top Ru(0001)-(2×1)-O hcp Ru(0001)-(2×2)3O hcp
5.87 4.82 4.23 5.52 5.29
Ru(0001)-(1×1)-O hcp O(bridge)-RuO2(110) O(3-fold)-RuO2(110) O(on-top)-RuO2(110) O2-RuO2(110)
5.07 4.60 5.80 3.20 0.80
a The internal O-O binding energy of O is calculated to be 6.0 2 eV. Except for the moelcular oxygen, all other energies are given with respect to atomic oxygen.
Figure 2. (a) Stick and ball model of the Oγ-RuO2(110) surface. Large balls represent oxygen, and small balls represent ruthenium atoms of RuO2(110). A bridge-bonded (Obr), an on-top bonded (Oγ), and a 3-fold coordinated O atom are indicated. (b) The optimum surface geometry of the Oγ-RuO2(110) as determined by LEED and DFT calculations. All values are given in angstroms. The corresponding layer spacings in bulk RuO2 are 1.27 and 0.635 Å. The Oγ-Ru bond length is 1.85 Å.
Figure 3. Comparison between experimental and calculated LEED IV curves for the optimum surface structure of the weakly held oxygen Oγ on RuO2(110). The agreement is quantified by RP ) 0.35.
by Oγ. Thereby, the underlying RuO2(110) surface undergoes only minor structural changes upon O adsorption. Most notably, however, the Rucus atoms move outward by 0.1 Å when capped by Oγ, to establish a bulklike environment. The adsorption of
Figure 4. RuO2(110) surface precovered by various coverages of Oγ and subsequently saturated by N2. N2 is known to adsorb directly above the Rucus atoms. If Oγ occupy the same adsorption site as N2, then the coverages of Oγ and N2 are related by θ(Oγ) ) θ(Rucus) - θ(N2). θ(Rucus) is the number of cus atoms on RuO2(110).
Oγ transforms the Rucus atom into a bulklike (i.e., 6-fold coordinated) Ru atom, as also corroborated by recent highresolution surface core level shift (SCLS) measurements.18 Similar structural results were also obtained by DFT calculations. The Ru-Oγ bond length turned out to be 1.79 Å, which is markedly shorter than the O-Ru bond length of chemisorbed atomic oxygen on Ru(0001) (2.00-2.03 Å)4 and in RuO2(110) (1.93 and 2.00 Å).8 This result is quite remarkable and may point to a coordination effect, since the binding energy of OγRucus is only 3.2 eV according to our DFT calculations and therefore significantly lower than that of the bridging Oγ-Ru bonding of 4.6 eV (cf. Table 1). The low adsorption energy makes Oγ a potentially active species. The adsorption site of Oγ can also be titrated by coadsorbing N2 onto RuO2(110). The basic property which allows this kind of titration experiment is that N2 adsorb above the Rucus atoms with their molecular axes normal to the surface plane.19 Therefore, Oγ and N2 will compete for the same adsorption sites, if Oγ adsorbs over Rucus atoms. In the N2 titration experiments, the RuO2(110) surface is precovered with various coverages of Oγ, and subsequently, the surface is saturated by N2. In the saturated N2-RuO2(110) overlayer all Rucus atoms are capped.19 With TDS both of the relative coverages of Oγ and N2 were measured. The titration experiments19 indicate that, with increasing Oγ coverage, the N2 coverage decreases in a way in which the total coverage of N2 and Oγ is preserved (cf. Figure 4). This finding provides additional evidence for the on-top adsorption of Oγ. At saturation of Oγ, still some N2 (20% of its saturation coverage) can adsorb on RuO2(110). This supports the above LEED analysis which indicates that only 75% of Rucus atoms are capped by Oγ at saturation.
Various Oxide Species on RuO2(110)
J. Phys. Chem. B, Vol. 105, No. 18, 2001 3755 TABLE 2: Vibrational Losses of O Chemisorbed on Ru(0001) and in RuO2(110)a
O species
HREELS: vibration (meV)
ref
low-coverage O Ru(0001)-(2×2)-O Ru(0001)-(2×1)-O Ru(0001)-(2×2)3O Ru(0001)-(1×1)-O O(bridge)-RuO2(110) O(on-top)-RuO2(110) O2-RuO2(110)
perp ) 54 perp ) 64 perp ) 73, lat ) 53 perp ) 80, lat ) 67 perp ) 81 perp ) 69 perp ) 103 O-O ) 142
30 44 38 38 30 this work this work this work
DFT: vibration (meV) 67.1 80.2 perp ) 63 perp ) 99 O-O ) 134
a ”Perp” and “lat” stand for the perpendicular and the lateral vibrations of oxygen, respectively.
Figure 5. Optimum surface geometry of the O2-RuO2(110) as determined by DFT calculations. O2 bridges two neighboring Rucus atoms. All values are given in angstroms. The internal O-O bond length is 1.30 Å, while the O-Ru bond length is 2.16 Å.
The DFT calculations predicted a molecular oxygen species Oδ on RuO2(110). The local adsorption geometry is defined by a flat lying oxygen molecule, which bridges two neighboring Rucus atoms (cf. Figure 5). The internal O-O bond length of this species is 1.30 Å, and the O-Ru bond length is 2.16 Å; note that the O-O bond length in a free O2 molecule is calculated to be 1.22 Å. In comparison to O2 on Pt(111),20,21 the structural properties of O2 on RuO2(110) are compatible with a superoxo-like O2- species. DFT calculations of O2 on Pt(111) revealed 1.34 Å for the O-O bond length and 2.07 Å for the Pt-O bond length.22 3.2. Energetics. Oxygen atoms chemisorbed on Ru(0001) (OR) form various ordered phases and desorb associatively between 1100 and 1700 K on heating the sample (cf. Figure 1). The heat of adsorption was roughly estimated to be 400 kJ/ mol (low coverage value) and 315 kJ/mol (0.5 monolayer), assuming second-order desorption.23 More recent values for the heat of adsorption as a function of O coverage were provided by a combined TDS/DFT investigation: 4.9 eV (very low O coverage), 4.25 eV (0.25 monolayer), 3.5 eV (0.5 monolayer), 3.0 eV (0.75 and 1.0 monolayer).24 At O coverages of 0.25, 0.5, 0.75, and 1 monolayer (2×2)-O, (2×1)-O, (2×2)-3O, and (1×1)-O overlayers are formed, respectively. Beyond an oxygen coverage of 1 monolayer, O2 desorption takes place at 1040 K; the corresponding desorption state is referred to as (Oβ) in Figure 1. As previously shown,5,8 excessive oxygen exposure to Ru(0001) at sample temperatures of 600-800 K leads to the epitaxial growth of ultrathin RuO2(110) domains, which may coexist with the (1×1)-O phase. Decomposition of this phase is accompanied by O2 desorption with a TDS peak at 1040 K (Oβ state). The bond strength of Oβ is thus considerably lower than for O atoms chemisorbed on Ru(0001) (OR). From TDS the bridging O species cannot be differentiated from the 3-fold coordinated lattice O of RuO2(110). The calculated binding energies (DFT) of oxygen on Ru(0001) as a function of the coverage are collected in Table 1 together with the bond strengths for the various oxygen species on RuO2(110). All DFT calculations were performed with the same program and the same pseudopotentials to ensure comparability. Clearly, the O-Ru bond strength of the O/Ru(0001) phases decreases by 0.8 eV when the coverage is increased from 0.25 to 1.0 monolayer; an identical trend was reported in a previous DFT study.25 The bridging O species of RuO2(110) is even more
weakly bound by 0.5 eV than O in Ru(0001)-(1×1)-O. The 3-fold coordinated lattice O atoms of RuO2(110) reveal a binding energy which is close to that of the (2×2)-O. However, if the bridging oxygen atoms are removed from the surface, then the thus generated 4-fold coordinated Ru atoms agglomerate in Ru clusters, thereby destabilizing the former 3-fold coordinated surface O atoms. This may explain why no extra desorption state is resolved in TDS for bridging O and 3-fold coordinated O on RuO2(110). It is quite interesting to note that O in the bridging position on the Ru(0001) surface, which is considered to be the transition state for the CO oxidation over Ru(0001),26 reveals a binding energy which is similar to that of Obr on RuO2(110) (cf. Table 1). Exposing the RuO2(110) surface to oxygen at room temperature leads to the population of a weakly held oxygen species (Figure 1, desorption state: Oγ) which desorbs in the temperature range of 400-500 K with second-order kinetics.6 Saturation of this desorption state occurs at an oxygen exposure of 5-10 L. The Oγ-Rucus binding energy is 3.2 eV and is therefore significantly lower than the bridging oxygen (cf. Table 1). It may be interesting to compare Oγ with the on-top O species identified on MoO3 27 and V2O5.28 Cluster calculations have shown that the bond strength (7.64 eV, 7.09 eV) of the on-top O species on MoO3 is higher than that of the highly coordinated O atoms (6.81 eV).27 Therefore, this O species is not considered to be particularly important for the activity of MoO3.29 In Table 1 we compare the binding energy of the chemisorbed oxygen atoms on Ru(0001) dependent on the coverage and the coordination number. Clearly, by going from the 3-fold hollow (hcp) to the bridge and finally to the on-top position, the adsorption energy changes by 0.8 and 1.6 eV, respectively. We may therefore conclude that the lower binding energy of the bridging O and the on-top O on RuO2(110) is mainly a consequence of the lower coordination. The molecular species Oδ on RuO2(110) was experimentally identified by a desorption state at about 140 K (cf. Figure 1). O2 on RuO2(110) is bound by 0.8 eV as estimated by our DFT calculations, thus being comparable in bond strength with O2 on Pt(111).21,22 3.3. Vibrations. For very low coverages of chemisorbed oxygen on Ru(0001), a vibrational loss at 54 meV is observed in HREELS, which was assigned to the perpendicular O against Ru vibration.30 With increasing O coverage, this stretching mode then shifts from 61 meV for the (2×2)-O to 81 meV for the (1×1)-O (cf. Table 2 for particular values). In Figure 6, we present HREEL spectra of the as-grown RuO2(110) surface in comparison with the Oγ covered RuO2(110) surface and the Ru(0001)-(1×1)-O. In the RuO2(110) related spectra, no loss at 81 meV is discernible, consistent with a complete RuO2(110) film covering the Ru(0001) surface. The oxide spectra are not
3756 J. Phys. Chem. B, Vol. 105, No. 18, 2001
Kim et al. TABLE 3: Work Function of Oxygen on Ru(0001) and RuO2(110) As Determined by DFT Calculations and Experimentsa work function (eV) system
DFT
expt
ref
Ru(0001) Ru(0001)-(2×2)-O Ru(0001)-(2×1)-O Ru(0001)-(2×2)-3O Ru(0001)-(1×1)-O RuO2(110) RuO2(110)-(1×1)-Oγ O2-RuO2(110)
5.03 5.42 6.07 6.54 6.74 5.90 7.40a 6.75
5.37 5.65 6.0 6.4 6.7 5.8b 6.6b
46 6 6 6 6 6 6
a The experimental coverage was only 75% of RuO (110)-(1×1)2 Oγ. b The Ru(0001) surface was not completely covered by RuO2(110).
Figure 6. High-resolution electron loss spectra for the clean RuO2(110) surface (a) and Oγ-RuO2(110) (b). The clean spectrum is dominated by a vibrational loss (69 meV) due to Obr against Ru. The additional loss at 103 meV in the Oγ-RuO2(110) is assigned to atomic oxygen in the terminal position above Rucus. Both assignments are based on DFT calculations. (c) Oγ-RuO2(110) after heating to 550 K in order to remove Oγ. (d) For comparison the vibrational spectrum of Ru(0001)-(1×1)-O is taken from ref 31. All spectra are recorded in specular geometry with a primary energy of 2.5 eV. The scaling factor in the loss region is 100.
disturbed by the appearance of strong Fuchs-Kliewer phonons due to the metallic character of RuO2(110). The clean RuO2(110) surface is dominated by a vibrational loss at 69 meV, which we attribute to the stretching mode of Obr against the Ru atoms underneath. This assignment is based on our DFT calculations that determined the Obr-Ru stretching mode on the clean RuO2(110) surface to be 63 meV in nice agreement with the experimental value. Other (but less strong) vibrational losses in Figure 6 are at 108 and 123 meV; the nature of these vibrational modes are unclear. The vibrational loss at 45 meV is in the range of vibrations, which were ascribed to RuO2 on the basis of surface enhanced Raman scattering (SERS) and Raman spectroscopy.31,32 It is important to recall that those investigations are performed under electrochemical conditions where the degree of order and crystallinity, in particular the exposed surface plane of RuO2, were largely unknown. In this context, we emphasize that under electrochemical conditions nanoclusters of RuO2(100) are grown on Ru(0001),33 while under ultrahigh vacuum conditions large areas of RuO2(110) are formed. This uncertainty may account for some of the deviations in the vibrational spectra obtained by SERS/Raman and HREELS. It is also worthwhile to mention that SERS and Raman investigations identified a vibrational loss at about 80 meV with RuO2(110).31,32 As evidenced by HREELS (cf. Figure 6), this vibrational mode is not associated with RuO2(110), but rather with the (1×1)-O overlayer, which may coexist with the oxide. If compared with the clean RuO2(110) surface, the Oγ covered RuO2(110) surface (cf. Figure 6) exhibits an additional loss at 103 meV in the HREEL spectrum, which is therefore related to Oγ. Flashing the Oγ covered RuO2(110) surface to 550 K removed the weakly held oxygen, as known from TDS, but also seen in HREELS (Figure 6, spectrum c). DFT calculations with supercell geometry determined for Oγ in on-top configuration an O-against-Ru stretching mode at 99 meV, which again compares favorably with the experimental value of 103 meV. In the literature31,32 this vibrational mode was erroneously assigned to RuO3. On the Oγ-RuO2(110) surface, the intensity of the Obr-against-Ru stretching mode is markedly reduced in comparison with the clean RuO2(110) surface (cf. Figure 6, spectra a + b). Since the Oγ atoms in on-top position stick
further out of the surface by about 0.7 Å than the Obr atoms, Oγ may screen the Obr-Ru vibration mode. The molecular oxygen species Oδ reveals a vibrational loss at 142 meV, which is assigned to the internal O-O stretching mode based on our DFT calculations (with a value of 134 meV). 3.4. Electronic Properties. According to our DFT calculations, the work function of the chemisorbed species OR on Ru(0001) increases from 5.03 to 6.74 eV when increasing the coverage from 0.25 to 1.0 monolayer (cf. Table 3). These values agree nicely with the experimental ones. The DFT calculated work function of the RuO2(110) surface is 5.9 eV, which rises to 7.40 eV when the surface is saturated by Oγ. Experimentally, the work function for a 6 monolayers precovered O-rich Ru(0001) phase is 5.8 eV and increases to 6.6 eV when the surface was saturated by Oγ.6 The deviation in the work function change between experiment and theory may be traced to the fact that the Ru(0001) surface was not completely covered with RuO2(110) in the experiments and that the saturation coverage of Oγ is 75% rather than 100% of the Rucus atoms as assumed in our DFT calculations. The strong increase in the work function with the population of Oγ is compatible with the ontop adsorption of Oγ, since the dipole length is large and the Oγ atom is strongly polarized as indicated by our DFT calculations (cf. Figure 7). In the literature of oxide surface chemistry nucleophilic and electrophilic oxygen atoms have been introduced to account for the activity and selectivity of particular reactions, such as the partial oxidation of methanol to formaldehyde or the ethylene epoxidation reaction. For the latter reaction, the concept of electrophilic oxygen (electron accepting, oxygen in a lower valence state) was introduced for Ag catalysts.3,34 To allow for a close approach of O to the CdC double bond, the electron cloud of electrophilic oxygen should be contracted. Actually, this requirement is met with the weakly held oxygen Oγ on RuO2(110) (cf. Figure 7). Density difference plots obtained by DFT calculations indicate that the charge cloud around Oγ is indeed more contracted than for Obr. Oγ on RuO2(110) may therefore represent a promising candidate for an electrophilic O species. At this point, a critical comment about electrophilic and nucleophilic oxygen has to be made. These concepts are adopted from molecular chemistry. In an oversimplified view, nucleophilic oxygen is identified with O2- and electrophilic oxygen with O- (e < 2). However, a direct correspondence to the charge state and electron distribution of oxygen is certainly not appropriate as exemplified with RuO2. Our DFT calculations estimated the charge transfer from Ru to oxygen in RuO2 to be only 0.3e so that none of the oxygen atoms in RuO2(110) would be nucleophilic in a strict sense.35 In contrast, a charge state of
Various Oxide Species on RuO2(110)
J. Phys. Chem. B, Vol. 105, No. 18, 2001 3757
Figure 7. Pseudovalence charge density contour plots of the Oγ-RuO2(110) surface cut through the Rucus atom along the [1h10] and the [001] directions from DFT calculations (left and middle). On the right side, the surface is cut through the bridging O atom to compare the charge density of Obr with that of O-ontop (left side). These plots are defined as the difference between the total valence electron density, as determined by DFT calculations, and a linear superposition of radially symmetric atomic charge densities. Contours of constant charge density are separated by 0.15 eV/Å3. Electron depletion and accumulation are marked by broken and solid lines. In addition, regions of charge accumulation are darker shadowed than those of charge depletion.
TABLE 4: Dissociative Sticking Coefficient of Oxygen on Ru(0001) and RuO2(110) O species
sticking coeff.
ref
initial Ru(0001)-(2×2)-O Ru(0001)-(2×)-O Ru(0001)-(2×2)-3O Ru(0001)-(1×1)-O O(on-top)-RuO2(110)
0.40 0.40 0.04 0.0003 at 600 K
